KR101879309B1 - Method and apparatus for extracting animate noun using possessive postposition - Google Patents

Abstract

A method for extracting a well name noun by using a tube survey according to an embodiment of the present invention includes a preprocessing step of reading text from an electronic document, a step of extracting a pronoun in the text and a cognate combined with a helping person, And if the first group of words is adjacent to the first word and the first group of words is adjacent to the first group, And extracting the first group of pronunciations into the well name nouns.

Description

[0001] METHOD AND APPARATUS FOR EXTRACTING ANIMATE NOUN USING POSSESSIVE POSTPOSITION [0002]

[0001] The present invention relates to a method and apparatus for extracting a well-known noun using a tube-type survey, and more particularly, to a method and apparatus for extracting a person's name from a Korean novel's text based on survey and tube- ) And a character (general noun) are extracted.

Information extraction is a task that extracts important information such as an object name (Named Entity) and an event from a natural language text. In particular, object name recognition is a part of information extraction, and after finding an object name in text, It is a technique to classify into a predefined class.

Most of these object recognition methods utilize rule - based algorithms or machine - learning based technologies. Recently, hybrid techniques have been proposed to reduce the disadvantages of the two techniques and utilize the advantages.

However, according to the analysis of object name recognition techniques, it is not easy to apply the existing object name recognition technique to a text genre other than the target text genre. However, Is limited to extracting the object name from the text such as a newspaper article without considering it. Most of the techniques proposed so far focus on extracting object names from text such as newspaper articles.

To this end, Patent Application No. 10-2016-0100737 discloses a method for extracting a well-known noun such as a character's name and a character using a plain-name use survey irrespective of the genre of the text, There is a problem that no well-known noun is not extracted. Here, a person's name means a proper noun such as "Harry" in the book "Harry Potter Series", and the character means a general noun such as "Prince" in the book "Little Prince".

A problem to be solved in the embodiment of the present invention is to provide a technique for extracting a well name noun such as a character name and a character using a relation with a well name noun extracted already for various genre texts without classifying the genre of the text.

In addition, in Patent Application No. 10-2016-0100737, in order to solve a problem in which it is not possible to extract a noun phrase that has not been used in conjunction with the use of a full name search, it is intended to provide a technique for extracting a noun phrase that is not used together with a full name use search .

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are intended to provide further explanation of the invention as claimed.

According to an embodiment of the present invention, there is provided a method for extracting a well name from a digital document, comprising the steps of: reading text from an electronic document; extracting a pronoun in the text and a cogniton combined with a helping person; Classifying a cognition combined with a usage survey into a first group, and if a noun phrase adjacent to a first word appears after a predetermined number of times in the text after a phrase in which a cognitive probe is combined with a cognition of the first group, Extracting the first group of pronunciations that are adjacent to the first word and are combined with the above-mentioned quadrangulation, and extracting the first group of pronunciations as wellnames.

At this time, the validity survey can be at least one of the composite surveys combined with -, -, -, and / or - with other surveys.

의 형태로 구성된 명사구 -상기

은 상기 제1 그룹의 체언이고, 상기

는 상기 관형격조사이며, 상기

은 상기 제1 단어임- 가 소정 횟수 이상 나타나는 경우, 상기 텍스트 내에서

의 형태로 구성된 명사구 -상기

는 상기 제1 그룹의 체언이 아니고, 상기

는 상기 관형격조사이며, 상기

은 상기 제1 단어임- 의 상기

를 유정명사로 추출하는 단계를 포함할 수 있다. In addition, the step of extracting with the well name noun

Noun phrase consisting of the form

Is a cognate of the first group,

Is the above-mentioned tube-type irradiation,

If the first word is a predetermined number of times or more,

Noun phrase consisting of the form

Is not a clan of the first group,

Is the above-mentioned tube-type irradiation,

Is the first word,

Into a well name noun.

A second group, a second group, a second group, a second group, a second group, a second group, a second group, a second group, a second group, The method according to claim 1, further comprising the step of classifying the voices not belonging to any of the third groups into a fourth group, The fourth group of cheongs are adjacent to the first group and are adjacent to the first group when the first group of words is adjacent to the first word after the first group of words is displayed a predetermined number of times or more in the text, And extracting a clan not belonging to the second group among the clan of the first group as a well name noun.

At this time, the amorphous use investigation is at least one of -, -, -, -, -, EO, -, -, - and - yen, and the general search can be at -.

의 형태로 구성된 명사구 - 상기

은 상기 제1 그룹의 체언 중 상기 제2 그룹 및 상기 제3 그룹에 속하지 않는 체언이고, 상기

는 상기 관형격조사이며, 상기

은 상기 제1 단어임- 가 소정 횟수 이상 나타나는 경우, 상기 텍스트 내에서

의 형태로 구성된 명사구 - 상기

는 상기 제4 그룹의 체언이고, 상기

는 상기 관형격조사이며, 상기

은 상기 제1 단어임 - 의 상기

를 유정명사로 추출하는 단계를 포함할 수 있다. In addition, the step of extracting with the well name noun

Noun phrase consisting of the form

Is a vowel not belonging to the second group and the third group of the first group of voices,

Is the above-mentioned tube-type irradiation,

If the first word is a predetermined number of times or more,

Noun phrase consisting of the form

Is a cognate of the fourth group,

Is the above-mentioned tube-type irradiation,

Is the first word,

Into a well name noun.

의 형태로 구성된 명사구 - 상기

이 상기 제1 그룹의 체언 중 상기 제2 그룹 및 상기 제3 그룹에 속하지 않는 체언이고, 상기

가 상기 관형격조사이며, 상기

이 상기 제1 단어임- 가 소정 횟수 이상 나타나는 경우, 상기 텍스트 내에서

형태로 구성된 명사구 - 상기

은 관형어가 아니고, 상기

는 상기 제4 그룹의 체언이며, 상기

는 상기 관형격조사이고, 상기

은 상기 관형격조사를 포함하지 않는 하나 이상의 관형어 또는 null이며, 상기

은 상기 제1 단어임 - 의 상기

를 유정명사로 추출하는 단계를 포함할 수 있다. In addition, the step of extracting with the well name noun

Noun phrase consisting of the form

Is not a member of the second group and the third group among the members of the first group,

Is the above-mentioned tube-type irradiation,

If the first word is a predetermined number of times or more,

Noun phrase consisting of

Is not a tongue,

Is a cognate of the fourth group,

Is the above-mentioned tube-type irradiation,

Is one or more idiomatic words that do not include the above-mentioned tube-type examination, or null,

Is the first word,

Into a well name noun.

In addition, the extracting step may include at least one of pronoun, unspecified noun, aggregate noun, plural noun, rhetorical noun, and dependent noun among the extracted good nouns.

Meanwhile, the preprocessing step includes dividing the text in quotation marks into utterances and classifying the texts other than the utterances into a narratuve, and the extracting of the voices includes extracting a phrase from the narrative, And classifying the words into a first group, wherein the classifying step includes classifying the words combined with the vocabulary usage search in the extracted vocabulary and the words combined with the vocabulary usage search in the utterance into the first group Step < / RTI >

Further, the step of extracting the cognition may include extracting words combined with - and -, and words combined with - and - with a cognate.

The apparatus for extracting a well name using a tube-type survey according to an embodiment of the present invention includes a preprocessor for reading text from an electronic document, a snippet extractor for extracting a snippet combined with a phrase search and assistant in the text, A group classification unit for classifying the voices combined with the vulgarity use search into a first group, and a grouping unit for classifying the voices that are adjacent to the first word in the text a predetermined number of times or more And a genuine noun extracting unit for extracting a genuine noun from the first group of pronouns.

According to the embodiment of the present invention, it is possible to extract a well name noun for various genre texts by extracting the well word noun using Korean features without limiting the genre of the text.

Also, it is possible to extract well nouns that are not used together with the survey of the use of the correct names by using the tube survey.

FIG. 1 is a functional block diagram of a well name extraction apparatus using a tube type survey according to an embodiment of the present invention.
FIG. 2 is a flowchart showing a process of a well-known noun extraction method using a tube-type survey according to an embodiment of the present invention.
FIG. 3A is a Venn diagram showing the relationship between the well and the amorphous nouns.
Figure 3b is a Venn diagram divided by a group of voices combined with a vulgar usage survey.
Figure 3c is a Venn diagram divided by a group of vignettes combined with a vulgar use survey, an amorphous usage survey, and a general purpose survey.
4 is a graph showing the number of words in a novel used to measure the accuracy rate and recall rate of the embodiments of the present invention.
FIG. 5 is a graph showing the accuracy rate of the well names extracted through the embodiments of the present invention with respect to the novel of FIG.
6 is a graph showing recall ratios of well names extracted through embodiments of the present invention with respect to the novel of FIG.
FIG. 7 is a graph showing the relationship between the recall rate and the recall rate when the target nouns having the appearance rates of the well names extracted through the embodiment of the present invention are 0%, 0.25%, 0.5%, 1% .

DETAILED DESCRIPTION OF THE EMBODIMENTS Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the like elements throughout. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment according to the present invention will be described in detail with reference to the accompanying drawings.

The embodiments disclosed herein should not be construed or interpreted as limiting the scope of the present invention. It will be apparent to those of ordinary skill in the art that the description including the embodiments of the present specification has various applications. Accordingly, any embodiment described in the Detailed Description of the Invention is illustrative for a better understanding of the invention and is not intended to limit the scope of the invention to embodiments.

The functional blocks shown in the drawings and described below are merely examples of possible implementations. In other implementations, other functional blocks may be used without departing from the spirit and scope of the following detailed description. Also, although one or more functional blocks of the present invention are represented as discrete blocks, one or more of the functional blocks of the present invention may be a combination of various hardware and software configurations that perform the same function.

In addition, the expression "including any element" is merely an expression of an open-ended expression, and is not to be construed as excluding the additional elements.

Further, when a component is referred to as being connected or connected to another component, it may be directly connected or connected to the other component, but it should be understood that there may be other components in between.

Also, the expressions such as 'first, second', etc. are used only to distinguish a plurality of configurations, and do not limit the order or other features between configurations.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a functional block diagram of a well name extraction apparatus 100 using a tube type survey according to an embodiment of the present invention. 1, a well name noun extracting apparatus 100 using a tube type survey according to an embodiment of the present invention includes a preprocessing unit 110, a snail extraction unit 120, a grouping unit 130, And an extracting unit 140.

)는 유정명사(

)와 무정명사(

)로 분류될 수 있으며, 이때 유정명사는 사람이나 동물 따위를, 무정명사는 식물이나 무생물을 가리키는 명사를 의미한다. 본 명세서에서 사용하는 유정명사는 사람이나 동물을 나타내는 일반명사와 고유명사를 포함하는 개념이다. 예를 들어, 유정명사 중 일반명사는 "아버지", "어머니" 등의 등장인물이 될 수 있고, 유정명사 중 고유명사는 "해리", "이사벨라", "빌리" 등의 인물명이 될 수 있다. The preprocessing unit 110 reads texts to be extracted from the electronic nouns. Referring to FIG. 3A, a noun used in one text (

) Is a well-known noun (

) And amorphous nouns (

), In which case the well name means a person or animal, and the amorphous name means a plant or a noun that refers to an inanimate object. As used in this specification, a well-known noun is a concept that includes a general noun and a proper noun representing a person or an animal. For example, a common noun can be a character such as "father", "mother", etc., and the proper name of a well name can be "Harry", "Isabella" or "Billy".

The snail extraction unit 120 extracts a snail from the text read by the preprocessing unit 110 by giving the snail and the snail associated with the auxiliary agent.

The grouping unit 130 classifies the cognitions combined with the vulgar name use search among the cognacs extracted by the vocabulary extraction unit 120 into the first group. The use of a vulgar name means an investigation that can be followed by a well name.

The well-known noun extracting unit 140 can extract the first group of pronunciations classified by the group classifying unit 130 as well names. In addition, even if the genuine noun extracting unit 140 does not belong to the first group, if the noun phrases adjacent to the specific pronoun appear in the text a predetermined number of times or more after the phrases in which the tube test is combined with the first group of pronouns, And can be extracted as a well-known noun. The tube survey refers to a survey of '-' s.

Hereinafter, a concrete procedure of the operation of each configuration of the well name extraction device 100 using the typing survey will be described with reference to FIG.

Meanwhile, the preprocessing unit 110, the snail extraction unit 120, the grouping unit 130, and the well name extraction unit 140, which are included in the above-described embodiment, include a memory including instructions programmed to perform these functions, And a microprocessor that executes these instructions.

FIG. 2 is a flowchart showing a process of a well-known noun extraction method using a tube-type survey according to an embodiment of the present invention. Each step of the method for extracting a well name using the method of FIG. 2 can be performed by the apparatus for extracting a name of a good 101 using the method shown in FIG. 1, and each step will be described below.

First, the preprocessing unit 110 reads text to be extracted from the electronic noun (step S210). The snail extraction unit 120 extracts a snail from the text read by the preprocessing unit 110 and extracts the snail associated with the auxiliary agent and the candidate as a candidate (S220).

In the Korean grammar, the nominative survey is defined as '-i / -', and the assistant is defined as '- / /', and '- / / is / / is / -' It is attached and used as. For example, "Harry said," and "Harry said." Therefore, it is possible to extract a used word from the read text by combining with at least one of '-i / - / - / / -'.

In addition, a cognate such as' Kok Sook 'with a foot on the last syllable can be combined with' - i 'and' - s', and a cognate such as' '. Thus, using these features, we can extract words combined with '-i' and '-u', and words combined with '-ga' and '-i'. However, the method for extracting a voice is not limited to the above-described example.

)으로 분류한다(S230). 유정명사용 조사란 유정명사 뒤에 붙을 수 있는 조사를 의미하며, 그 예는 아래 표 1과 같다.3B, the grouping unit 130 searches the first group (the first group) for the first group (step S 220)

) (S230). Useful name survey refers to an investigation that can be followed by a well name. Examples are shown in Table 1 below.

At this time, the basic type of the use survey is' - to / to / / - 'and it is possible to classify the voices combined with the basic type of the voiced use survey as the first group, and the' , And '-' to '-' to '-' and '-'.

Next, the well name noun extracting unit 140 can extract a cognate belonging to the first group as a well name noun (S240). However, if the first group of pronouns are directly extracted as a well-known name, a large number of the extracted good nouns can be pronouns, nouns that represent an unspecified person, or an investigation or dependent noun.

Accordingly, in step S240,

1) pronouns (eg, 'me', 'us', 'him', 'her', etc.)

2) Unspecified nouns (eg, 'people', 'men', 'women', etc.)

3) Collective nouns (eg, 'family', 'family', 'crowd', etc.)

4) Plural form (eg 'people', 'men', 'women', etc.)

5) Investigation (eg, 'one', 'two', 'three', etc.)

6) It is possible to extract well nouns except for cases where they depend on dependent nouns (eg 'noun', 'name', 'minute', etc.).

In this way, the well name noun extracting unit 140 can extract a cognate except a pronoun, an unspecified noun, an aggregate noun, a plural form, an investigation, and a dependent noun among the first group of nouns.

On the other hand, there may be a well-known noun, even if it is not a cognate combined with a vulgarity usage survey in text. In this case, in order to extract a well name noun, in step S240, the well word noun extracting unit 140 extracts a specific noun phrase from the first group noun phrase If a noun phrase adjacent to a cognate appears more than a predetermined number of times in the text, a cognate that is adjacent to a certain cognate and combined with the tube search can be extracted as a well name noun.

이 제1 그룹의 체언이고,

가 관형격조사이며,

이 특정 체언인 명사구가 텍스트 내에서 소정 횟수 이상 나타나는 경우, 아래 [실시예 2]와 같이

가 제1 그룹의 체언이 아니고,

가 관형격조사이며,

이 [실시예 1] 의 체언인 명사구에서

를 유정명사로 추출할 수 있다.For example, in step S240, the well name extraction unit 140 extracts

This is the first group of voices,

Of the patients,

In the case where the specific noun phrase appears in the text for a predetermined number of times or more, as in the case of [Example 2]

Is not the first group's voice,

Of the patients,

In the cognate phrase of this [Example 1]

Can be extracted as a well-known noun.

)의(

) 목소리(

)는", "철수(

)의(

) 목소리(

)가", "민수(

)의(

) 목소리(

)를"과 같은 형태의 문장이 텍스트에서 나타나는 경우 (즉, 3명 이상의 유정명사의 관형격조사 다음에 나타나는 경우), "해리(

)의(

) 목소리(

)는"과 같은 문장이 텍스트에서 검색된다면 '해리'는 유정명사용 조사와 결합하여 쓰이지 않았다고 하더라도 유정명사로서 추출할 수 있다. 이는 유정명사에 관형격조사가 결합된 어절의 뒤에 위치하는 체언은, 작가 또는 저자에 의해 다른 유정명사에 관형격조사가 결합된 어절의 뒤에 사용될 가능성이 높기 때문이다. For example, in the text, 'Young-hee', 'Chul-soo', and 'Min-su'

)of(

) voice(

) "," Withdraw (

)of(

) voice(

), "" Min-su (

)of(

) voice(

) "Appears in the text (ie, if it appears after the tube survey of three or more well-known nouns)," Harry

)of(

) voice(

) Can be extracted as a well name even if it is not used in conjunction with the use of a valid name search if the sentence such as "is searched in the text." This is because the word " Or it is likely that it will be used by authors behind other vernacular verses that incorporate a tube survey.

On the other hand, among the voices extracted by using only the voiced name use survey, it is possible to include the amorphous nouns due to the anthropomorphism, etc. When the amount of the text is large, the voices combined with the tube type survey are searched for all the voices not belonging to the first group Is inefficient, the method of extracting a well name using a tube survey according to an embodiment of the present invention can classify a group of the cognos more subdivided and classify the group of the cognos that are not the proper nouns, for example, After excluding the group of censors combined with the group and the general survey, the remaining pronouns can be extracted based on the tube survey.

Figure 3c is a Venn diagram divided by a group of vignettes combined with a vulgar use survey, an amorphous usage survey, and a general purpose survey.

), 무정명사용 조사와 결합된 체언을 제2 그룹(

) 및 범용 조사와 결합된 체언을 제3 그룹(

)으로 분류할 수 있고, 제1 그룹, 제2 그룹 및 제3 그룹의 어디에도 속하지 않는 체언을 제4 그룹으로 분류하는 단계를 포함할 수 있다. 이때 무정명사용 조사는 무정명사에만 결합되는 조사로 '-에까지', '-에는', '-에도', '-에로', '-에서', '-에서는', '-에선' 및 '-엔' 을 포함하고, 범용조사는 유정명사 및 무정명사에 모두 결합될 수 있는 조사로 '-에'를 포함한다. 2 and 3C, in step S230, the group classifier 130 classifies the voices combined with the strong name use search among the voices extracted by the voicing extractor 120 into the first group (

), A cognate combined with an amorphous use survey into a second group (

) And the cognition combined with the general survey were assigned to the third group (

), And classifying the voices not belonging to any of the first group, the second group and the third group into the fourth group. At this time, the amorphous use survey is an investigation that is combined only with amorphous nouns, and it is an investigation that is combined with '- to', 'to', 'to', 'to', 'to', 'to', ' Yen ', and the general purpose survey includes'-e' as an investigation that can be combined with both well and noun.

At this time, the amorphous noun can be combined with the unbiased search using the technique of personification, objectification, etc., and the well name noun may be combined with the amorphous use search according to the intention of the author or author. ) Can be extracted as a well-known noun excluding the second group or third group of the first group's voices, and the second group or the third group, , The fourth group of pronunciations adjacent to the specific pronunciated and combined with the annotation survey can be extracted as the wellnote nouns when the noun phrase adjacent to the specific annotation appears more than a predetermined number of times in the text. It can also exclude pronouns, unspecified nouns, aggregated nouns, plural forms, rhetoricals, and dependent nouns.

이 제1 그룹의 체언 중 제2 그룹 및 제3 그룹에 속하지 않는 체언이고,

가 관형격조사이며,

이 특정 체언인 명사구가 텍스트 내에서 소정 횟수 이상 나타나는 경우, 아래 [실시예 4]와 같이

이 제4 그룹의 체언이고,

가 관형격조사이며,

가 [실시예 3]의 체언인 명사구에서

를 유정명사로 추출할 수 있다. For example, in step S240, the well name extraction unit 140 extracts

Is not a member of the second group and the third group of the first group,

Of the patients,

In the case where the phrase of the specific cognate appears in the text a predetermined number of times or more, as in the case of [Example 4]

This is the fourth group of voices,

Of the patients,

Quot; is the cognate word of " Example 3 "

Can be extracted as a well-known noun.

을 아래 [표 2]에 개시된 조사(

)와 결합된 체언으로 한정할 수도 있다. [표 2]는 국어국립원 표준국어대사전에 포함된 366개의 조사 중에서 인물명의 뒤에 붙을 수 있는 조사와, 발화에서 인물명이 불릴 때 사용되는 기호, 및 이들의 조합을 선정한 것이다. 또한 추출한 유정명사 중 대명사, 불특정명사, 집합명사, 복수형, 수사 및 의존명사를 제외할 수도 있다.At this time, in [Example 3] and [Example 4]

(See Table 2 below)

). ≪ / RTI > [Table 2] is a survey of 366 surveys included in the National Language Standard Dictionary of Korean National Standard Language, which is used when a person name is called in utterance, and combinations thereof. It is also possible to exclude pronouns, unspecified nouns, aggregated nouns, plural forms, rhetoric, and dependent nouns.

이 제1 그룹의 체언 중 제2 그룹 및 제3 그룹에 속하지 않는 체언이며,

가 관형격조사이고,

이 특정 체언인 형태의 명사구가 텍스트 내에서 소정 횟수 이상 나타나는 경우, 아래 [실시예 6]과 같이

이 관형어가 아니거나 null이고,

가 제4 그룹의 체언이며,

가 관형격조사이고,

이 관형격조사를 포함하지 않는 하나 이상의 관형어이거나 null이며,

이 [실시예 5]에서 추출된 체언인 명사구에서

를 유정명사로 추출할 수 있다. In addition, in step S240, the well name extraction unit 140 extracts

Are not included in the second group and the third group of the first group,

Is a tube-

In the case where the noun phrase of the specific cuneo form appears in the text a predetermined number of times or more,

Is not an idiomatic word, or is null,

Is the fourth group's vowel,

Is a tube-

This is one or more idiomatic words that do not include the tube test, or null,

In the cognate noun phrase extracted in [Example 5]

Can be extracted as a well-known noun.

이 관형어인 경우에는

을 유정명사로 추출하지 않을 수 있다. 가령, "또 한(

) 명의(

) 얼굴(

)이 자신을 바라보고 있었다, (

은 null)" 의 경우와 같이 의존명사('명')는 주로 관형어('한')가 수식하는 경우가 많기 때문이다. In this case, in [Embodiment 6], since the idiomatic expresses mainly amorphous characters, etc.,

In the case of this tongue

Can not be extracted as a well-known noun. For example,

) Name(

) Face(

) Was looking at herself,

As in the case of "null"), dependence nouns ('noun') are mostly modifiers ('han').

와

사이에는

을 수식하는 단어(

)가 위치할 수 있는데, 이때

이 소유격인 경우에는 정확률을 향상시키기 위해

을 유정명사로 추출하지 않을 수 있다. 가령, "사막의 하늘에서 떨어지던 생명의 양식인(

) 만나(

)의(

) 최초의(

) 이름(

)은 ..."에서, '만나'는 관형어인

에 의해서도 유정명사로 추출되지 않지만,

가 없다고 하더라도 '최초의(

)'라는 소유격의 존재 때문에 유정명사로 추출하지 않는다.In [Example 6]

Wow

Between

The word (

) Can be located,

In the case of this possessive, to improve accuracy

Can not be extracted as a well-known noun. For example, "the form of life that fell from the sky of the desert

) meet(

)of(

) First(

) name(

) Is ... "," meet "is an idiotic

Is not extracted as a well name,

Even if there is no 'first (

) 'Because of the existence of possessive do not extract as a well.

을 위 [표 2]에 개시된 조사(

)와 결합된 체언으로 한정할 수도 있다.On the other hand, in [Example 5] and [Example 6]

Of the survey (Table 2)

). ≪ / RTI >

Meanwhile, in step S210 of the method of extracting a genuine noun by using the tube-type survey according to an embodiment of the present invention, the preprocessing unit 110 may divide the text into Utterance and Narrative.

An utterance means that the idea of a novel character is realized in units of sentences, and the artist uses quotation marks ("...", "...") to indicate that a particular sentence is a utterance. Narrative is a set of sentences that lead to the story of a novel, consisting of sentences describing the narrative of a series of events in terms of first person or third person. To this end, the preprocessing unit 110 may classify the text in the quotation marks by speech, and divide the other text into narratives.

Although the well nouns appear as subjects in the narrative, they may not be extracted even by utilizing the nouns and assistants, because the nouns are not given as a subject or replaced with pronouns by omission. For this reason, when extracting a cognition, only the text corresponding to the narrative is extracted and the cognition is extracted. Then, when extracting the well name noun with the use of the full name use survey, the whole noun phrase extracted from the narrative, Can be extracted.

On the other hand, another embodiment of the present invention is a method for extracting a well name using a connection search.

Each step of the method for extracting a well name using a connection search according to an embodiment of the present invention can be performed through the apparatus for extracting a well name of FIG. 1, and each step will be described below.

First, the preprocessing unit 110 reads the text to be extracted from the wellhead nouns from the electronic document. The snail extraction unit 120 extracts the snail from the text read by the preprocessing unit 110 by giving the snail and the snail associated with the auxiliary agent.

)으로 분류한다. 3B, the grouping unit 130 classifies the cognition combined with the name use search among the cognate extracted by the cognate extracting unit into the first group (

).

Thereafter, the well name noun extracting unit 140 can extract a cognate belonging to the first group as a well name noun. However, if the first group of pronouns are directly extracted as the canonical name, a great number of the extracted canonical nouns may be pronouns, irrelevant nouns representing irrelevant persons, Can be extracted as genetic nouns except for pronouns, unspecified nouns, aggregated nouns, plural forms, investigations, and dependent nouns.

On the other hand, there may be a well-known noun, even if it is not a cognate combined with a vulgarity usage survey in text. In this case, for the extraction of the well word nouns, the well word noun extracting unit 140 extracts the words of the first group, which are not included in the first group but are combined with the connection probes ('-and' and '-and' Can be extracted as a well name noun.

이 제1 그룹의 체언이고,

가 접속 조사이면,

을 유정명사로 추출할 수 있다.For example, when the sentence included in the text has the same structure as that of the seventh embodiment described below, the well-known noun extracting unit 140 extracts,

This is the first group of voices,

Is a connection check,

Can be extracted as a well-known noun.

)와(

) 철수(

)는 같은 반이다" 형태의 문장이 있는 경우, '철수'는 '영희와'라는 어절과 인접하는 체언이므로 '철수'라는 체언이 전체 텍스트에서 유정명사용 조사와 결합되어 사용되지 않았다고 하더라도 유정명사로 추출할 수 있다. 반대로 [실시예 7]의 문장에서

가 접속 조사이고,

이 제1 그룹의 체언이면

을 유정명사로 추출할 수 있다. 이는 유정명사가 접속조사와 쓰이는 경우 대부분 다른 유정명사와 연결되어 사용되기 때문이다. For example, in the text "

)Wow(

) Withdrawal

) Is the same half. "If there is a sentence of the form," Cheolsu "is adjacent to the word" Yeonghee ", so even if Cheongsu is not used in the whole text in conjunction with the investigation of the use of vulgarity, On the other hand, in the sentence of [Example 7]

Lt; / RTI >

If this is the first group of words

Can be extracted as a well-known noun. This is because, when a well name is used for connection survey, it is often used in connection with other well names.

On the other hand, among the voices extracted using only the voiced name use survey, the voiced names may be included due to an anthropomorphism or the like, and when the amount of text is large, the voices combined with the connection search are searched for all the voices not belonging to the first group Therefore, the method of extracting a well name using a connection search according to an embodiment of the present invention can classify a group of the cognos more subdivided and classify the group of the cognos, which is not a proper noun, for example, After excluding the group of censors combined with the group and the general survey, the remaining ness can be extracted based on the connection survey.

), 무정명사용 조사와 결합된 체언을 제2 그룹(

), 범용 조사와 결합된 체언을 제3 그룹(

)으로 분류할 수 있고, 제1 그룹, 제2 그룹 및 제3 그룹의 어디에도 속하지 않는 체언을 제4 그룹으로 분류하는 단계를 포함할 수 있다. 이때 무정명사용 조사는 무정명사에만 결합되는 조사로 '-에까지', '-에는', '-에도', '-에로', '-에서', '-에서는', '-에선' 및 '-엔' 을 포함하고, 범용조사는 유정명사 및 무정명사에 모두 결합될 수 있는 조사로 '-에'를 포함한다. As shown in FIG. 3C, the group classifying unit 130 classifies the cognition combined with the vulgar name use search among the cognos extracted by the clan extractor 120 into the first group (

), A cognate combined with an amorphous use survey into a second group (

), The cognition combined with the general survey was assigned to the third group (

), And classifying the voices not belonging to any of the first group, the second group and the third group into the fourth group. At this time, the amorphous use survey is an investigation that is combined only with amorphous nouns, and it is an investigation that is combined with '- to', 'to', 'to', 'to', 'to', 'to', ' Yen ', and the general purpose survey includes'-e' as an investigation that can be combined with both well and noun.

At this time, due to the technique of personification, objectification, etc., the amorphous noun can be combined with the unbiased search and the well word noun can be combined with the amorphous usage search. Therefore, the well word noun extracting unit 140 extracts the second Group or third group can be extracted as a well-known noun except for the first group, the second group, or the third group, except for the third group, And extracting the group pronunciations into the wellnames. It can also exclude pronouns, unspecified nouns, aggregated nouns, plural forms, rhetoricals, and dependent nouns.

가 관형어가 아니거나 null이고,

이 제4 그룹의 체언이며,

가 접속조사이고,

이 제1 그룹의 체언 중 제2 그룹 및 제3 그룹에 속하지 않는 체언이면,

을 유정명사로 추출할 수 있다. For example, when the sentence included in the text has the same structure as that of the eighth embodiment below, the well name extraction unit 140 extracts,

Is not an idiomatic word or null,

This is the fourth group of voices,

Lt; / RTI >

If the first and second groups do not belong to the second group and the third group,

Can be extracted as a well-known noun.

가 관형어인 경우에는

를 유정명사로 추출하지 않는다. 가령, '민수가 좋아하는(

) 음식(

)과(

) 민지(

)가(

) 사랑하는 음식은 비슷하다'의 경우, '음식'는 관형어 '좋아하는'의 수식을 받는 무정명사이므로 추출하지 않는다. At this time, amnesty is often used in conjunction with a tongue

Is a tongue-shaped word

Are not extracted as well names. For example,

) food(

)and(

) Minji(

)end(

) In the case of 'the food of love is similar', 'food' does not extract because it is an amorphous noun to receive the expression of 'liking'.

을 유정명사 중 인물명으로 한정할 수 있으며, 위 [표 2]에 개시된 조사(

)와 결합하여 사용된 체언을 유정명사로 추출할 수 있다. On the other hand, in [Example 8]

Can be limited to a person's name among the well names, and the survey (Table 2)

) Can be used to extract the vowel used as a well name.

이 제1 그룹의 체언 중 제2 그룹 및 제3 그룹에 속하지 않는 체언이고,

가 접속조사이며,

이 제4 그룹의 체언이고,

가 접속조사 및 목적격 조사가 아닌 경우,

을 유정명사로 추출할 수 있다. In addition, if the sentence included in the text has the same structure as the [example 9] below, the well name extraction unit 140 extracts,

Is not a member of the second group and the third group of the first group,

≪ / RTI >

This is the fourth group of voices,

Is not a connection survey or an objective investigation,

Can be extracted as a well-known noun.

가 접속조사 및 목적격 조사('-을', '-를')인 경우에는

을 유정명사로 추출하지 않는다. 가령, '빌리는 해리와(

) 피자를(

) 먹는다'의 경우와 같이

가 목적격 조사인 경우에는 무정명사와 결합될 수 있기 때문이고, '나는 철수(

)와(

) 피자(

)와(

) 콜라를 먹었다'에서'피자'의 경우에는 '철수'와 동격으로 사용된 것이 아니라 뒤따라오는 '콜라'와 동격으로 사용된 것일 수 있기 때문에 제외하도록 한다.At this time, in [Example 9]

('-', '-'),

Are not extracted as well names. For example, "Billy said,

) Pizza (

) As in the case of eating

Is the subject investigation, it can be combined with the amorphous noun.

)Wow(

) Pizza(

)Wow(

) In the case of 'eaten cola' to 'pizza', it is not used as an equivalent to 'Cheol-su', but it should be excluded because it may be used as a 'cola' following.

가 관형어 또는 형용사이면

을 유정명사로 추출하지 않을 수 있다. '많'라는 단어는 체언은 아니지만, '많+은'및 '많+이'와 같은 형태로 많이 사용되고 있기 때문에 체언으로 추출될 수도 있다. 가령, '영희(

)와(

) 많(

)은(

) 이야기를 나누었다'와 같은 경우이다. 따라서 오류를 줄이기 위해

가 관형어 또는 형용사이면

을 유정명사로 추출하지 않는다. [Example 9]

Is an adjective or an adjective

Can not be extracted as a well-known noun. The word 'many' is not a vowel, but it can be extracted as a cognate because it is often used in the form of 'many + silver' and 'many + i'. For example,

)Wow(

) Many

)silver(

) I have shared a story. So to reduce the error

Is an adjective or an adjective

Are not extracted as well names.

을 위 [표 2]에 개시된 조사(

)와 결합하여 사용된 체언을 유정명사로 추출할 수 있다. 또한 추출한 유정명사 중 대명사, 불특정명사, 집합명사, 복수형, 수사 및 의존명사를 제외할 수도 있다.In addition, in [Example 9]

Of the survey (Table 2)

) Can be used to extract the vowel used as a well name. It is also possible to exclude pronouns, unspecified nouns, aggregated nouns, plural forms, rhetoric, and dependent nouns.

FIGS. 4 to 7 are graphs for explaining the results of extracting the well names from the embodiments of the present invention.

For the calculation of the accuracy rate and recall rate for the result of the extraction of the well nouns according to the above-described embodiment, the extracts of the extracted well nouns and those corresponding to the actual noun nouns were manually extracted and compared.

The accuracy and recall rate are defined as follows.

The accuracy rate is calculated as the ratio of actual good nouns excluding errors among the good nouns extracted according to the embodiment, with respect to the extracted good nouns according to the embodiment. The recall rate is calculated as the actual person name ratio among the well names extracted through the embodiment, with respect to the actual person names (corresponding to proper nouns among the well names extracted) manually extracted.

4 is a graph showing the number of words in a novel used to measure the accuracy rate and recall rate of the embodiments of the present invention.

Referring to FIG. 4, this experiment was conducted in 80 Korean novels, and novels having about 100,000 words were used in the experiments. The x-axis of FIG. 4 represents a novel index, which is a novel in which novels 1 to 11 are translated into Korean, and novels 12 to 80 are Korean.

FIG. 5 is a graph showing the accuracy rate of the well names extracted through the embodiments of the present invention with respect to the novel of FIG. 5 is an example of extracting a well name noun by using only a valid name use survey in an AP, and an example in which a well word noun is additionally extracted using a connection survey in a CFoAN, and an example in which a well word noun is additionally extracted using a tube survey, , The connection investigation and the tube survey are all used to extract the oil well nouns is called CPFoAN.

According to the results shown in FIG. 5, a total of 1811 well names were extracted by AP from all 80 books. Of these, 1,776 were correctly extracted, and the accuracy rate of AP was 98.07%. A total of 1,882 well names were extracted by CFoAN The accuracy of CFoAN was 98.51%. A total of 2,075 well names were extracted by PFoAN. Of these, 2,037 were correctly extracted. The accuracy rate of PFoAN was 98.17%. A total of 2,128 wells Nouns were extracted, but 2,082 of them were correctly extracted, so the accuracy rate of CPFoAN is 97.84%.

6 is a graph showing recall ratios of well names extracted through embodiments of the present invention with respect to the novel of FIG. 6 is an embodiment for extracting a well name noun using only the use of a full name use survey as an AP, and an example for extracting a well name noun as an addition survey using a connection survey as an example of a CFoAN, , The CPFoAN is an example of additionally extracting the well names using both the connection survey and the tube survey.

According to the results of FIG. 6, a total of 1007 person names are extracted by the AP in a total of 804 volumes having a total number of 1443 persons, the recall rate of the AP is 70.27%, a total of 1078 persons are extracted by CFoAN, 75.23%. The total recall of AP1 is 81.79% and the total recall of AP is 93.95% by CPFoAN.

FIG. 7 is a graph showing the relationship between the recall rate and the recall rate when the target nouns having the appearance rates of the well names extracted through the embodiment of the present invention are 0%, 0.25%, 0.5%, 1% .

The rate of appearance is calculated as the ratio of the frequency of appearance of one person to the sum of the frequency of appearance of all the well names. For example, the rate of occurrence of A is 1%, which means that the rate of appearance frequency of A is 1% with respect to the sum of frequency of all the wells in the novel.

Therefore, if we set the target rate of 0% or more as the target, we will search all the nouns of all the wells, and we can confirm that the recall rate is even higher even if the target rate is set to 1% It is possible to easily extract the noun propery through the embodiment of the present invention.

From the above results, it can be seen that the dictionary names can be effectively extracted for various texts without discrimination in the genre of the text through the above-described embodiment. Furthermore, in addition to the use of the use of the full name use survey, the full use rate and the recall rate can be increased by extracting the full name use survey and the unused full name noun by using the quadratic survey, so that the above embodiment can more efficiently extract the well noun.

The above-described embodiments of the present invention can be implemented by various means. For example, embodiments of the present invention may be implemented by hardware, firmware, software, or a combination thereof.

In the case of hardware implementation, the method according to embodiments of the present invention may be implemented in one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs) , FPGAs (Field Programmable Gate Arrays), processors, controllers, microcontrollers, microprocessors, and the like.

In the case of an implementation by firmware or software, the method according to embodiments of the present invention may be implemented in the form of a module, a procedure or a function for performing the functions or operations described above. The software code can be stored in a memory unit and driven by the processor. The memory unit may be located inside or outside the processor, and may exchange data with the processor by various well-known means.

Thus, those skilled in the art will appreciate that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. It is therefore to be understood that the embodiments described above are to be considered in all respects only as illustrative and not restrictive. The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

100: Well-known noun extraction device using a tube survey
110:
120:
130:
140: Well noun extraction unit

Claims (13)

A preprocessing step of reading text from an electronic document;
Extracting a pronoun in the text and a vocabulary associated with the assistant;
Classifying the extracted voices into a first group; And
When the first word is adjacent to the first word and the first word is adjacent to the first word in the first group, the first word is adjacent to the first word and the first word is adjacent to the first word, And extracting the first group of pronunciations as candidates of the wellnot nouns
Well - known noun extraction method.
The method according to claim 1,
The validation survey should include at least one of the combined surveys combined with other surveys to - / to / / - and / to - / to /
Well - known noun extraction method.
The method according to claim 1,
Wherein the step of extracting candidates of the well names includes:
Within the text

Noun phrase consisting of the form

Is a cognate of the first group,

Is the above-mentioned tube-type irradiation,

If the first word is a predetermined number of times or more,
Within the text

Noun phrase consisting of the form

Is not a clan of the first group,

Is the above-mentioned tube-type irradiation,

Is the first word,

As candidates of the well name nouns
Well - known noun extraction method.
The method according to claim 1,
A first group, a second group, a second group, a second group, a second group, a third group, a second group, a second group, and a third group, Further comprising the step of classifying the voices not belonging to any of the third group into the fourth group,
Wherein the step of extracting candidates of the well names includes:
When a first word and a noun phrase adjacent to the first word appear after a predetermined number of times in the text after the first and second words of the first group and the words of the first group are not included in the second group and the third group, Extracting a fourth sentence of the fourth group, which is adjacent to the front of the word and is combined with the above-mentioned survey of the fourth group, and the second sentence of the second group,
Well - known noun extraction method.
5. The method of claim 4,
The amorphous use investigation is at least one of - to, to, - to, - to, - to, - to, - to,
The general purpose irradiation is -
Well - known noun extraction method.
5. The method of claim 4,
Wherein the step of extracting candidates of the well names includes:
Within the text

Noun phrase consisting of the form

Is a vowel not belonging to the second group and the third group of the first group of voices,

Is the above-mentioned tube-type irradiation,

If the first word is a predetermined number of times or more,
Within the text

Noun phrase consisting of the form

Is a cognate of the fourth group,

Is the above-mentioned tube-type irradiation,

Is the first word,

As candidates of the well name nouns
Well - known noun extraction method.
5. The method of claim 4,
Wherein the step of extracting candidates of the well names includes:
Within the text

Noun phrase consisting of the form

Is not a member of the second group and the third group among the members of the first group,

Is the above-mentioned tube-type irradiation,

If the first word is a predetermined number of times or more,
Within the text

Noun phrase consisting of

Is not a tongue,

Is a cognate of the fourth group,

Is the above-mentioned tube-type irradiation,

Is one or more idiomatic words that do not include the above-mentioned tube-type examination, or null,

Is the first word,

As candidates of the well name nouns
Well - known noun extraction method.
The method according to claim 1,
Wherein the step of extracting candidates of the well names includes:
Further comprising the step of excluding at least one of the extracted candidates, an unspecified noun, an aggregate noun, a plural form, an investigation, and a dependent noun
Well - known noun extraction method.
The method according to claim 1,
The pre-
Classifying text in quotation marks by utterance and classifying the non-speech text into a narrative,
The method according to claim 1,
Extracting a pronoun in the narrative and a cognac associated with the assistant;
Wherein the grouping into the first group comprises:
Categorizing words extracted from the vocabulary into a first group and vocabulary combined with a vocabulary usage search and a word combined with a vocabulary usage search in the utterance
Well - known noun extraction method.
The method according to claim 1,
The method according to claim 1,
- < / RTI > and < RTI ID = 0.0 > - < / RTI &
Well - known noun extraction method.
A preprocessing unit for reading text from an electronic document;
A snippet extracting unit for extracting a snail from the text and a snail associated with the assistant;
A grouping unit for classifying the extracted voices into voiced first group, And
When the first word is adjacent to the first word and the first word is adjacent to the first word in the first group, the first word is adjacent to the first word and the first word is adjacent to the first word, And a well name noun extracting unit for extracting the first group of pronouns as candidates of the well name nouns
Well - known noun extraction device.
A program stored on a computer readable medium for causing a processor to perform the method of any one of claims 1 to 10.
11. A computer-readable medium having stored thereon instructions for causing a processor to perform the method of any one of claims 1 to 10.

Images